Research report · Run 8 · 28 September 2026
Monthly cross-channel briefing: a job to investigate
Decision: search for a narrowly scoped offline SKILL.md candidate. Do not add a seventh catalogue card or recommend an install. No reporting package was sought or inspected in this run.
What prompted the question
Report format and interpretation
A February 2026 agency question asks how to combine Google Ads, GA4, and Meta reporting, and whether to use platform reports, a dashboard, slides, or a short summary. Replies differ, but several emphasize changes and next actions. This is a qualitative report-design signal, not a representative preference study. Some replies promote tools.
Manual assembly, but demand for automation
A March 2026 practitioner post describes assembling Search Console, GA4, and rank figures into a Google Doc. It is an unverified time self-report. The author wants automatic pulling and branded delivery across clients—work the proposed offline skill would not do.
A short narrative can matter
In an August 2026 agency discussion, one respondent says a short note with key figures, changes, and next steps was read more than automated reports; another values a monthly client call. The initiating product-research post was removed. These are self-selected anecdotes, not measured read or retention rates.
Connected reporting is an alternative
A January 2026 agency thread includes an operator using connected AgencyAnalytics reports without AI. AgencyAnalytics' own 2026 guide presents templates, dashboards, and automated delivery. These are user self-reports and vendor sales material, not independent prevalence or time-saving evidence. They show that this job is not unserved.
Inference, not finding: the possible gap is a repeatable briefing and review protocol for someone who already has approved exports, not a connector or dashboard. A generic prompt or template may be enough; a portable skill would need to add inspectable inputs, source-to-claim mapping, arithmetic checks, discrepancy stops, and human approval.
A first-use brief for a later package search
1 · Authorized input
One organization; one closed current period and a comparison period with timezone and extraction date; at least two manually exported aggregate channel summaries; original source, metric label, unit and definition for every row; approved goals or targets; and an owner-written log of campaigns, outages, tracking changes, and caveats. The owner chooses the recipient and question. No personal records, credentials, hidden dashboards, proprietary margins, or raw event logs.
2 · Reviewable draft
One local Markdown or document brief: period and cutoff, a compact channel/source table, change summary, observations separated from possible explanations, owner-confirmed context, missing data and uncertainty, two or three human-owned next questions, and a mapping from every number and claim to a supplied row. Use “unknown” rather than inventing figures. It is not an addressed email or scheduled PDF.
3 · Calculation stop
Normalize units, currencies, windows, and scope before comparison. Difference = current − prior. Percentage change = (current − prior) ÷ prior × 100 only when prior is nonzero. Build a rate from a numerator and denominator with the same source, definition, and period. Do not average rates or sum differently attributed conversions. Recalculate a blended paid metric only after the owner approves the numerator, denominator, and deduplication definition.
4 · Human sign-off
Flag zeros, missing rows, changed tracking, delayed periods, and discontinuities. The authorized account lead or in-house owner checks rights and sensitivity, each figure against the export, definitions, attribution, caveats, factual and causal claims, brand/legal suitability, next actions, recipient, and exact version. Only that person circulates it. Draft approval does not authorize account access, refresh, publication, sending, spend, or campaign changes.
Google Ads discrepancy guidance cites conversion delay, lookback/count/attribution settings, and processing as reasons reports can differ. GA4 modeling documentation says channel-attributed conversion data may be updated for up to 12 days. These are product-specific cautions, not a rule that one platform is correct. A numerical change does not establish why it occurred or that marketing caused it.
How it differs from the six current entries
| Existing entry | Overlap | Missing from that job |
|---|---|---|
| Copywriting | Could polish approved wording. | Metric definitions, period comparisons, reconciliation, sourced monthly report. |
| Content Strategy | Could turn approved findings into a later plan. | Monthly channel-export analysis and report. |
| Competitor Discovery | Could supply separately dated public context. | Client channel results and performance claims. |
| Search Page Audit | An audit could be a reported activity. | Evidence that search performance changed, or a recurring cross-channel brief. |
| Ad Campaign Analyzer | Closest overlap: pasted or exported ad metrics and recommendations. | Reconciliation across channels and definitions, stakeholder narrative, period/source notes, and distribution sign-off. Its recommendations do not authorize account action. |
emails | Email activity might be one aggregate row. | Email performance measurement or a report; the package is a conditional draft-only candidate with no send route. |
This comparison uses the existing six candidate descriptions and Run 7. It is not a fresh package inspection or a current-branch verification.
Search boundary and rejection test
A spreadsheet, slide template, document template, or generic prompt may already solve the job with less setup. Looker Studio and connected reporting platforms are better fits when a team primarily needs live data, automatic collection, or scheduled delivery. The Ad Campaign Analyzer source is relevant to paid-export analysis, not enough on its own for a cross-channel stakeholder brief.
Defer or reject this skill job if follow-up interviews or a small workflow test find no recurring reconciliation or narrative bottleneck for people with approved exports; if an existing template or prompt produces an adequately sourced, reviewed brief; if manual export is unacceptable; or if useful reporting requires sensitive data, credentials, automatic delivery, or unsupported causal claims. A later search can also fail if no original-source SKILL.md fits the bounded input → check → draft → approval path, or if provenance or access is too uncertain. No candidate has yet been evaluated against these tests.
Checked 28 September 2026 UTC. Evidence includes the four community discussions and vendor guide linked above, product-specific Google documentation, the existing public candidate, Run 7, and the pool's audience and runner notes. No first-party analytics, interviews, user test, package test, outcome measurement, time-saving claim, or market-prevalence estimate is available. Research library · Six-entry catalogue